Organizations have been trying to apply artificial intelligence (AI) to the task of records management for more than 20 years, mostly without success. But it’s worth giving this technology another look.
The tools needed to mine through vast quantities of data in order to find patterns have advanced by leaps and bounds in just the last five years. A sub-category of artificial intelligence, known as machine learning, is the most promising of these.
One of the primary inhibitors to the use of AI in records management has been the tedious and time-consuming process of training the computer algorithms to look for specific elements. Machine learning does away with much of this drudgery.
Machine learning algorithms can comb through unstructured text and learn about the format and the content as they go. For example, teaching a computer to translate phrases from English to French no longer requires a massive dictionary. Instead, the machine can infer basic structure and meaning from just a few sentences. As it learns, additional data can be fed into it in order to enhance its understanding of the languages. The process is automatic and requires almost no human intervention.
This same process can be applied to one of the most arduous and error-prone areas of records management: classification. Giving a computer a few documents that have been pre-classified by humans provides enough information to get it started on larger sets of records. Humans are needed to confirm the machine’s work in the beginning, but one of the benefits of machine intelligence is that the computer gets “smarter” as it goes along — eventually eliminating the need for any human intervention.
Machine learning can also be used to weed out documents that are no longer needed or are duplicates. These big data-optimized devices excel at poring over millions of files. It’s a simple matter for them to identify, flag and even automatically delete duplicates. The machines can also detect things like date fields for use in records retention. Google already does this in presenting the most recent information first. The same thing can be done with corporate records.
Another potential application of AI is compliance. Organizations often have only a few days to assemble all of the documents needed for an audit. It’s typically a frantic and error-prone process for humans, but machines can be trained to spot keywords, labels or patterns that identify a document as relevant for compliance purposes, then pull them out of a repository in minutes.
Finally, AI can assist with data quality, which is a problem that afflicts every organization. Data entered by humans is prone to formatting errors, misspellings, misplaced information and other glitches. This can cause duplication issues and hinder the ability of organizations to get a full view of their critical data. With machine learning, computers can be trained to look for something like ZIP codes that are entered into date fields or records that have the same address but different names. The devices can automatically correct many of these errors, enabling organizations to significantly improve the quality of their data.
Some people worry that artificial intelligence will put employees out of work, but it’s likely that most individuals would be happy to be rid of the tasks described above and it would allow them to dedicate more of their workday to higher-value objectives. AI and machine learning have finally moved beyond the realm of hypothetical potential to real-world application.
If this topic interests you, and you’re planning to attend the ARMA Live! conference in Orlando next month, be sure to stop by “Robots Meet Records: Using Artificial Intelligence in IG and Privacy” on Monday afternoon to learn more.